One giant leap for AI: Companies rethink how and where data is processed

The Orbital Edge: Why the Future of AI is Moving into Space

For decades, satellites have functioned primarily as “bent pipes”—collecting data and beaming it back to Earth for processing. However, as the AI-driven data economy expands, this model is hitting a wall. The sheer volume of data produced by modern earth observation satellites is overwhelming available bandwidth, creating a bottleneck that slows down real-time decision-making.

The solution? Moving the “brain” of the operation into orbit. By implementing edge computing in space, the industry is shifting toward orbital data centers that process information at the source, transmitting only the most critical insights back to ground stations.

Solving the Latency Crisis with Space-Based AI

In sectors where seconds matter—such as disaster response, border surveillance, and defense—waiting for data to travel from a satellite to a ground station and back can be a critical failure. Processing data at the edge in space allows for quicker insights and enhanced mission autonomy.

Anirudh Sharma, CEO of Digantara, notes that edge computing is essential for reducing downlink and information latency. Beyond simple data transmission, this capability enables onboard inference. For example, satellites within a constellation can exchange data via inter-satellite links to maintain the constellation and avoid collisions without needing a command from Earth.

This autonomy becomes even more vital in higher orbits, such as Geostationary (GEO) and beyond. In these environments, “ground-in-the-loop” decision cycles are extremely demanding, making onboard autonomy the primary infrastructure for effective decision-making.

Turning Satellites into Intelligent Nodes

The integration of machine learning (ML) is transforming satellites from passive sensors into intelligent nodes. AI models now allow satellites to prioritize, compress, and interpret high-value data despite the strict power and compute limitations of the space environment.

Selective Transmission and Intelligent Filtering

The goal is not to process everything in orbit, but to make smarter decisions about what actually needs to be sent home. Awais Ahmed, founder and CEO of Pixxel, explains that the real value lies in “filtering, intelligent compression, or prioritising what to transmit first.”

Pixxel already utilizes these techniques for cloud detection and compression to optimize how data is transmitted. By moving intelligence closer to the source, companies can improve responsiveness even as still relying on ground infrastructure for deeper, large-scale model execution.

Data-Centre-Class Computing in Orbit

The ambition extends beyond simple filtering. The Spaceborne Computer programme by Hewlett Packard Enterprise (HPE) demonstrates that data-centre-class computing can be extended into space.

HPE’s Spaceborne Computer-2, currently aboard the International Space Station, integrates high-performance computing (HPC) and AI using commercial off-the-shelf hardware. Ryan D’Souza, HPE’s country manager for AI and HPC, suggests that for deep-space or lunar missions—such as those led by the Indian Space Research Organisation (ISRO)—near real-time data analysis at the edge can significantly boost operational efficiency.

Strategic Applications: Beyond Earth Observation

The move toward orbital processing is creating a ripple effect across multiple global industries:

• Defence and Intelligence (ISR): Real-time tracking of adversary satellite movements and space debris. Digantara, for instance, aims to deploy a constellation of 15 satellites for space domain awareness by 2027.
• Climate and Agriculture: Rapid monitoring of crop health or climate shifts without the lag of traditional downlink cycles.
• Disaster Management: Immediate identification of flood or fire zones to trigger emergency responses in minutes rather than hours.

Pawan Kumar Chandana, CEO of Skyroot Aerospace, emphasizes that because space compute qualifies as critical infrastructure, the ability to process data directly in orbit is a matter of national and operational sovereignty.

The Engineering Hurdle: Power, Thermal, and Reliability

Despite the potential, building a data center in a vacuum is not simple. Engineers face a “trilemma” of constraints: power, thermal management, and reliability.

AI inference in orbit must operate within incredibly tight margins. Processing generates heat, and in the vacuum of space, dissipating that heat is a major challenge. The reliability of onboard analysis is paramount; as Anirudh Sharma points out, ensuring there are no “false positives” is a critical constraint for high-stakes decision-making.

most experts agree that space-based computing will complement, rather than replace, terrestrial infrastructure. While the “first-order” decisions (filtering and prioritizing) happen in orbit, the “deep analytics” will remain grounded.

Frequently Asked Questions